Okay, I am back after a bit of playing with the Legos. I had two identical motors hooked up to the same power source. One motor was directly driving one side of the differential. The other motor was geared
down 24:17 (or 17:24 depending on how you look at it). First, I ran one side at a time with the other side motor disconected and axle locked in place. I felt with my fingers the torque of the system and listened to the pitch of the sound to indicate speed. Then, I ran the system with both motors together. The speed was more than that of the geared down motor alone but less than the direct drive motor alone. The torque of the system was more than either motor alone.
So, my guess is that this design does work. The speed is averaged and the torque is added.
As for Andy Baker's post above, perhaps use a planetary differential instead like
this one
EDIT:
The way that one in the pic works is the top left and bottom right planetary gears are in contact with the sun gear which is attatched to the output on one side. All the planetary gears are in contact with each other. There is another sun gear that is not pictured that comes into contact with the bottom left and top right gears in the pic. That second sun gear is on the output shaft on the other side. I hope that helps explain it. I have the unit opened up here on my desk if you have any questions.